2x1=10

Let’s assume you already have an image in numpy’s ndarray for­mat, e.g. because you loaded it with OpenCV’s imread() func­tion, and you want to con­vert it to TensorFlow’s Ten­sor for­mat and lat­er back to ndarray.

That’s essen­tial­ly three calls to Ten­sor­Flow:

import cv2
import tensorflow as tf
import numpy as np

# normalize the pixel values to 0..1 range and convert them 
# to a single-precision tensor
image_in = cv2.imread('image.png') / 255.
t = tf.convert_to_tensor(image_in, dtype=tf.float32)
assert isinstance(t, tf.Tensor)

# in order to convert the tensor back to an array, we need
# to evaluate it; for this, we need a session
with tf.Session() as sess:
    image_out = sess.run(fetches=t)
    assert isinstance(image_out, np.ndarray)

# for imshow to work, the image needs to be in 0..1 range
# whenever it is a float; that's why we normalized it.
cv2.imshow('Image', image_out)
cv2.readKey(0)

Note that instead of using sess.run(t) we could also have used

with tf.Session() as sess:
    image_out = t.eval(sess)

which essen­tial­ly per­forms the same action. A ben­e­fit of using sess.run() direct­ly is that we can fetch more than one ten­sor in the same pass through the (sub-)graph (say, tuple = sess.run(fetches=[t1, t2, t3])), where­as call­ing tensor.eval() always results in one sep­a­rate pass per call.